overprovisioning
also: resource overcommitment, overcommitment
Allocating more resources (CPU, memory, storage, or network bandwidth) to virtual machines or containers than physically available, relying on the assumption that not all will be used simultaneously.
Overprovisioning is a resource management strategy where a hypervisor or container orchestrator assigns more total resources to guests than exist on the host system. For example, a physical server with 16 GB of RAM might be provisioned to run 10 virtual machines, each with 2 GB allocated, totaling 20 GB—exceeding actual capacity.
This works because most workloads don't use their full allocation at once. During low-demand periods, idle resources can be shared across multiple guests. However, if all systems request their resources simultaneously, contention occurs, leading to performance degradation or out-of-memory conditions.
Overprovisioning increases density and cost efficiency but requires careful monitoring. Systems like KVM, Xen, and Kubernetes often use memory ballooning, page sharing, or swap to handle peaks. It's commonly paired with resource limits and Quality of Service (QoS) policies to prevent one workload from starving others.